pci: vfio: Filter out some PCI extended capabilities

There are PCI extended capabilities that can't be passed through the VM
as they would be unusable from a guest perspective. That's why we
introduce a way to patch what is returned to the guest when the PCI
configuration space is accessed. The list of patches is created from the
parsing of the extended capabilities in that case, and particularly
based on the presence of the SRIOV and Resizable BAR capabilities.

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Signed-off-by: Steven Dake <sdake@lambdal.com>

pci/src/vfio_user.rs

@@ -11,6 +11,7 @@ use crate::{
 use anyhow::anyhow;
 use hypervisor::HypervisorVmError;
 use std::any::Any;
+use std::collections::HashMap;
 use std::os::unix::prelude::AsRawFd;
 use std::ptr::null_mut;
 use std::sync::{Arc, Barrier, Mutex};
@@ -112,6 +113,7 @@ impl VfioUserPciDevice {
             msi_interrupt_manager,
             legacy_interrupt_group,
             vfio_wrapper: Arc::new(vfio_wrapper) as Arc<dyn Vfio>,
+            patches: HashMap::new(),
         };
 
         // No need to parse capabilities from the device if on the restore path.